Analyses were performed in SPSS 26.0 (IBM Corp, Armonk, NY, USA). Firstly we constructed a directed acyclic graph to establish possible confounders and mediators (Figure S1). Then we explored differences between women exposed to episiotomy or not in the first birth by VE using test of proportions, based on the directed acyclic graph. We considered differences with a p‐value of <0.10 as possible confounders since these covariates had a 90% likelihood or more of being associated with the outcome (Table 1). Factors analyzed but omitted from Table 1 due to nonsignificant differences were smoking, cohabitation, hypertension and preeclampsia (data not presented).
Secondly, we assessed risk factors for prelabor cesarean delivery in the second birth using covariates from Table 1 in univariate and multivariate regression models. Results are presented as crude and adjusted odds ratios (aOR) with 95% confidence intervals (95% CI) in Table 2. The multivariate model included all factors significant in the univariate analysis: maternal age, country of birth, education, gestational age, epidural, labor dystocia, intrapartum fetal distress, station, head position, head circumference, birthweight, shoulder dystocia, Apgar 1 min <4, Apgar 1 min <7, episiotomy, year of delivery and region of delivery.
Thirdly, the propensity score (the conditional probability of being assigned episiotomy or not) was calculated using all covariates with a p‐value of <0.10 in Table 1: maternal age, country of birth, maternal height, maternal BMI, higher education, gestational age, epidural, labor dystocia, intrapartum fetal distress, fetal station, head position, head circumference, birthweight, shoulder dystocia, Apgar 1 min <4, Apgar 1 min <7, year of delivery and region of delivery. The propensity score was then used to perform a regression analysis and to calculate an inversed probability of treatment (episiotomy) weight for each individual as initially described by Rosenbaum.18 We used a modified computer syntax for SPSS provided by Thoemmes et al.19 The weight was used to account for bias due to observed confounders creating a pseudo‐population in which the covariates and the treatment assignment (episiotomy or not) are independent of each other, to mimic a randomized treatment assignment.20, 21, 22 We assessed the outcome using all obtained stabilized weights, as well as truncated stabilized weights, at the 5th and 95th percentiles or the 1st and 99th percentiles.
Fourthly, since episiotomy and OASIS are associated, we explored the prevalence and association of prelabor cesarean delivery in the second birth in women with four principal groups of exposure: “neither episiotomy nor OASIS”, “episiotomy, no OASIS”, “OASIS, no episiotomy” and “both OASIS and episiotomy”, using “neither episiotomy nor OASIS” as reference. The association was tested using multivariate logistic regression adjusting for maternal age, country of birth, higher education, gestational age, epidural, labor dystocia, intrapartum fetal distress, station, head position, head circumference, birthweight, shoulder dystocia, Apgar at 1 min, year of delivery, and region of delivery. Moreover, interaction between episiotomy and OASIS was formally tested using multivariate logistic regression entering the interaction term “episiotomy*OASIS”, “episiotomy”, “OASIS”, and all the confounders used in the multivariate model.
Brismar Wendel S., Liu C, & Stephansson O. (2023). The association between episiotomy or OASIS at vacuum extraction in nulliparous women and subsequent prelabor cesarean delivery: A nationwide observational study. Acta Obstetricia et Gynecologica Scandinavica, 102(3), 378-388.